It turns out that you need to add an argument to the test executable.
Under the Debug Configurations for the test you need to set a program argument under the Arguments tab.
--catch_system_errors=no
Otherwise the executable ignores the information about the breakpoint.